match point for Scarlett Johansson. The actress yesterday opened a serious debate about the ethical limits of artificial intelligence on the slippery slope of using familiar voices in AI chatbots. OpenAI announced on Monday that it was retiring Sky, the female voice that debuted the abilities of its new GPT-4o language model in the spoken version of ChatGPT last week. During Monday’s presentation, the great ability to talk about AI in an empathetic way and his vocal tone were reminiscent of the movie Her, in which a man (Joaquin Phoenix) falls in love with an intelligent artificial intelligence who is voiced by, in the original English version, Johansson.
Most of the media – including La Vanguardia – made reference to the film to report on the presentation of OpenAI. It is not accidental, because the similarities were sought. A note published yesterday by the actress revealed that the technology company wanted to hire her to use her voice, but she turned down the offer “for personal reasons”.
OpenAI CEO Sam Altman contacted Scarlett Johansson in September to sign her on. “He told me that he believed that if I put a voice in the system, I could build a bridge between tech companies and creatives and help consumers get comfortable with the seismic shift around humans and AI. He told me that he believed that my voice would comfort people”, explained the actress.
After rejecting the offer, Scarlett Johansson did not hear anything about it again, according to her version, until two days before GPT-4o was presented Altman contacted the actress’ agent to ask – him to reconsider, but there were no negotiations. Before there were new contacts, GPT-4o arrived, a high-capacity model that reacts to conversations with human-like speed.
“When I heard the demo – said Johansson – I was stunned, furious and in disbelief that Mr. Altman had searched for a voice that sounded so strangely similar to mine that neither my closest friends nor the media could notice the difference”.
According to Johansson, “Altman even hinted that the resemblance was intentional, tweeting a single word, ‘her,’ a reference to the film in which I voice a chat system, Samantha, who strikes up a relationship intimate with a human”.
Johansson has hired legal counsel and requested by letter Altman and OpenAI to detail “the exact process through which they had created the Sky voice.”
When denouncing the situation, the actress raised the flag against voice and image cloning with which generative AI threatens to flood the creation of audiovisual content. “At a time when we are all faced with deepfakes – highly likely false copies – and the protection of our own image, our work and our identity, I believe that these issues deserve absolute clarity.”
Now, Scarlett Johansson says she expects “a resolution in the form of transparency and the passage of appropriate legislation that helps ensure the protection of individual rights.” In October, the actress sued another AI company that copied her appearance and voice for an advertisement of her abilities in an ad published on X.
Had the clone of the actress’ voice been used without permission, OpenAI would have violated the executive order signed last fall by US President Joe Biden to prevent some misuses of artificial intelligence.
OpenAI responded to the controversy by publishing a text on its blog stating that Sky’s voice “is not an imitation of Scarlett Johansson, but belongs to a different professional actress who uses her own natural voice” . “To protect your privacy – added the company – we cannot share the names of our announcers”.
The company says it published a call for voice selection a year ago, to which more than 400 proposals were submitted, before choosing the five voices that were finally selected. One of these voices was Sky, precisely the one used in the presentation of the new GPT-4o. “We believe that AI voices should not deliberately imitate the distinctive voice of a celebrity: Sky’s voice is not an imitation of Scarlett Johansson, but belongs to a different professional actress who uses her own natural voice ,” OpenAI stated.
Along these lines, he also detailed that each selected actor flew to San Francisco between June and July last year for the recording sessions and that their voices were released on ChatGPT in September.
The controversial suspicion of possible misuse of Johansson’s voice comes at a time when Altman faces other ethical challenges, including the loss of OpenAI’s security team after its chief , Jan Leike, announced his departure from the company last week. This official accused the CEO of prioritizing the launch of new products to the detriment of their security. Along with Leike and other members of his team, Ilya Sutskever, one of the company’s co-founders and chief scientist, has also left OpenAI.
Three weeks ago, the US Department of Homeland Security announced the creation of the Artificial Intelligence Security Board to advise the Secretary of State and other stakeholders on the safe development and deployment of AI in infrastructure . Altman was first on the list.